Automated Identification and Classification of Brain Tumors Using Hybrid Machine Learning Models and MRI Imaging

Abstract

The need for automated diagnostic systems in medical imaging, particularly in the detection and categorization of brain tumors, is paramount. This research proposes a hybrid model to identify and classify MRI-detected brain tumors into four categories: pituitary, meningioma, glioma, or absence of a tumor. This hybrid approach leverages the strengths of both deep learning and traditional machine learning techniques, enabling the extraction of complex features and the recognition of intricate patterns, such as those found in brain tumors. Machine learning further enhances the model's capacity to classify accurately based on these specific features, reducing time and cost. The proposed system consists of several stages: initial pre-processing of brain MRI images, the application of two distinct segmentation techniques (region-based and edge-based), morphological operations, feature extraction, and finally classification. The classification employs a hybrid model (VGG16) in conjunction with four traditional classifiers: Support Vector Machine (SVM), Naive Bayes (NB), Decision Tree (DT), and Random Forest (RF). The experimental results highlight that the use of Random Forest with region-based segmentation yields the highest accuracy, reaching 99.17%. This combination excels at focusing on minute yet crucial details in MRI images and maintains stability in the presence of distortion and outliers. The dataset employed in this study is an amalgamation of three: Figshare, SARTAJ, and Br35H, each containing MRI images of the aforementioned four types of brain tumors.